摘要
为解决实际应用中彩色图像配准问题,针对已有的超复数互相关方法没有办法处理存在粗差的情况,在超复数互相关方法的基础上,结合鲁棒核函数,提出了一种超复数鲁棒相关的方法.该方法在频域中能使用快速傅里叶变换实现,在满足鲁棒性、实时性和准确性的同时,能通过遍历搜索去获得图像配准的全局解.即使在配准图像间存在粗差或者色彩旋转的情况下,该方法也能获得正确的配准结果.实验结果证明了该方法的有效性.
A new approach that incorporates hypercomplex correlation with robusmess is presented for color image registration. Existing hypercomplex correlation method is not robust against outliers. The proposed method utilizes advantage of the FFT, therefore it can be robust, fast and exhaustive. The method also performs well even outliers or rotation in 3-D space exist. The experimental results verify the effectiveness of the proposed method.
出处
《复旦学报(自然科学版)》
CAS
CSCD
北大核心
2007年第1期91-95,共5页
Journal of Fudan University:Natural Science
关键词
图像配准
超复数
互相关
鲁棒性
registration
hypereomplex
correlation
robust